Theoretical modelling was undertaken to determine the desired composition of well to achieve 2µm emission. A MATLAB®based finite well Schr¨odinger solver was used to find the energies of the carriers within the wells and subsequently the emission wave- length. The effective masses, band gaps and alignments were calculated using a Excel spreadsheet containing the equations outlined in Section 2.4. The equations incorporated strain and band bowing parameters taken from the comprehensive work of I. Vurgaftman
et al. [132] and material data from the Ioffe Institute [45].
Modelling suggested that quantum wells with an indium content of 35% would emit light in the target region around 2µm without reaching the strain limit and with suitable carrier confinement. To validate this a series of samples were grown on n-GaAs(100) substrates, this was facilitated by the introduction of an IMF. The IMF was achieved by firstly removing the oxide layer from the substrate. 100 nm of GaAs was then grown at 1 Monolayer per second (ML/s) with a substrate temperature of 575. Gallium flux was
then interrupted and the sample was left in a excess of arsenic for 1 minute. The arsenic flux was then switched off for 15 s allowing arsenic atoms to desorb, leaving the growth terminated on gallium atoms. This process was confirmed through observation of the RHEED pattern as it shifted from an As-rich (2×4) to a Ga-rich (4×2). This pause without flux, allowed for an abrupt gallium terminated interface to be achieved without As-Sb intermixing. The antimony valve was then opened for 4 minutes as the substrate was cooled to 515. Once the substrate temperature had stabilised the gallium shutter was opened and the growth continued with 400 nm of GaSb to create a buffer layer. The quantum wells were grown onto this, with a composition of Ga0.65In0.35Sb and well
widths as outlined in Table 5.1.
The quantum wells were grown at a rate of 1 ML/s. The composition was confirmed by the deposition of a bulk layer on a separate test wafer which was analysed by XRD. A 50 nm barrier of GaSb was grown between each quantum well. So that the GaSb cell temperature remained constant throughout the growth, the barriers were deposited at a rate of 0.65 ML/s. The quantum wells were then capped with 10 nm of GaSb.
X-ray diffraction (HRXRD) scans were taken of each sample to confirm the structures grew to the intended design and to give an indication of the structural quality. Two examples are shown in Figure 5.1. All samples except QA339 showed pendell¨osung fringes. Though they are less well defined than those simulated, this is consistent with XRD of other reported IMF samples[130].
Sample Well Thickness (nm)
QA276 6 QA277 4.8 QA280 3.6 QA333 4 QA338 8 QA339 12
Table 5.1: Well dimensions of Ga0.65In0.35Sb quantum well samples grown using IMF
(a) (b)
Figure 5.1: (a) XRD of QA280 consisting of 5×3.6 nm Ga0.65In0.35Sb QWs of pen-
dell¨osung fringes consistent with a periodicity of 51 nm (b) XRD of QA339 consisting of 5×12 nm Ga0.65In0.35Sb QWs evidence of relaxation with several additional features
seen. Theω−2θ axis of both samples has been zeroed to GaSb.
The periodicity of each samples’ quantum wells was calculated from the fringes using the equation[125];
t= (i−j)λ 2 (sinωi−ωj)
(5.1) where t is the thickness of a combined well and barrier layer, (i−j) is the number of fringe peaks between peaksiand j and sinω is the angle of peaks. λis the wavelength of the x-rays (0.1154 nm).
The resultant periods were consistently 2-3 nm smaller than the designed structures. The offset does not scale with well width, so is attributed to a thinner barrier of around 47 nm thick. It is not believed that this will compromise the results as it is still greater than the carrier wavefunctions penetration into the barriers. XRD simulations of QA339 (Figure 5.1b) indicate that the sample has relaxed as there are no clear fringes. Trans- mission electron microscopy (TEM) could have been used to determine if this is the case, but the weak photoluminescence of the sample led to it not being considered viable and further analysis excessive. The ω−2θ axis on each graph was zeroed to allow ease of
Figure 5.2: 4 K photoluminscence of Ga0.65In0.35Sb QW samples grown on IMF, showing
a clear red-shift with increasing well thickness. Additional peaks are due to contamina- tion in the system and not inherent to the QWs.
comparison with future GaSb substrates samples. The full width half maxima (FWHM) of the GaSb peak can be used to indicate the crystalline quality of the samples. Larger FWHM indicate poorer quality. As the well width increased the FWHM also increased. Photoluminescence measurements were conducted with laser power densities of up to 10 W/cm2 on the sample. At temperatures from 4 K until they quenched. The 4 K spectra have been normalised in Figure 5.2. Relative quality can be gained through FWHM of the spectra and the intensity of the signal. Additional peaks in Figure 5.2 for QA280, QA333 and QA338 have been attributed the contamination in the cryostat as they remained constant over several different samples, including the ones consisting of other material systems.
The 4 K photoluminescence emission peak wavelengths are shown in Figure 5.3a and are consistent with the MATLAB® simulations for all samples except QA339. This further suggests that it is beyond the critical thickness and no longer has abrupt wells. Comparing the quenching temperature of the samples in Figure 5.3b shows it increas- ing for well widths up to 6 nm at which point a decrease in quenching temperature is
(a) (b)
Figure 5.3: (a) A comparison of the 4 K photoluminescence emission peak wavelength with the modelled values. The dashed line indicates the nextnano® simulated peak emission of Ga0.65In0.35Sb wells at 4 K. (b) Figures of merit for the PL spectra of
Ga0.65In0.35Sb QW samples with integrated intensity of emission in black and tempera-
ture of PL quenching in red.
observed. Below 6 nm the reduced quenching temperature is attributed to decreasing confinement, the energy levels are higher in the wells due to the effect of the barrier potential. Above 6 nm the increased defects reduce the radiative efficiency and subse- quently the quenching temperature. The superiority of the 6 nm sample is also replicated in the integrated intensity. The full width half maxima (FWHM, ∆E) of the 4 K PL scan can be used to further compare the inhomogeneity using the equation[133].
∆E= ~ 2 m∗ e ∆LZ L3 Z (5.2)
where m∗e is the effective mass of the electron in the well, ∆LZ is the quantum well
interface roughness andLZ is the quantum well width.
The interface roughness increases with well width (Table 5.2), indicating that the strain is affecting the interfaces. Further properties of the 6 nm sample QA276, including the activation energy and recombination processes were also examined. These will be discussed in the next section.
Sample Well Thickness (nm) ∆LZ (ML) QA276 6 1 QA277 4.8 1 QA280 3.6 0.5 QA333 4 3 QA338 8 7 QA339 12 30
Table 5.2: Composition of Ga0.65In0.35Sb QW samples grown on IMF and ∆LZ calcu-
lated from the broadening of the 4 K photoluminescence spectra